New MPIA Working Paper - COnnecting REpositories · 2017. 5. 5. · IDRC photo: N. McKee MPIA...
Transcript of New MPIA Working Paper - COnnecting REpositories · 2017. 5. 5. · IDRC photo: N. McKee MPIA...
s
IDRC photo: N. McKee
M P I A W o r k i n g P a p e r 2006-06
María Inés Terra Marisa Bucheli Silvia Laens Carmen Estrades
April 2006
María Inés Terra (Universidad de la República, Uruguay) [email protected] Marisa Bucheli (Universidad de la República, Uruguay) [email protected] Silvia Laens (CINVE, Uruguay) [email protected] Carmen Estrades (Universidad de la República, Uruguay) [email protected]
The Effects of Increasing Openness and Integration to the MERCOSUR on the Uruguayan Labour Market: A CGE Modelling Analysis
This work was carried out with the aid of financial and scientific support from the Poverty and Economic Policy (PEP) Research Network (www.pep-net.org), financed by the International Development Research Centre (IDRC). The authors acknowledge the collaboration of Gabriel Katz in different phases of the study. We are very grateful to Maurizio Bussolo, Nitesh Sahay, Bernard Decaluwe, Nabil Annabi, John Cockburn, Randy Spence and Alvaro Forteza for their useful comments and suggestions. All remaining errors and omissions are our own responsibility.
Abstract
Uruguay is a small economy. Its integration into MERCOSUR has increased its exposure to regional macroeconomic instability. The aim of this paper is to assess the impact of regional integration on the country’s labour market and poverty. We estimated wage differentials between labour categories, finding a 60 percent wage gap between formal and informal workers. A CGE model with an efficiency wage specification for unskilled labour was built, with results showing that regional shocks deeply affect the Uruguayan economy. The consideration of an efficiency wage model is particularly important when shocks lead to a reallocation of resources towards sectors intensive in unskilled labour. A subsidy on formal, unskilled labour could contribute to decrease informality and therefore increase GDP, but this type of policy needs to be carefully implemented because it may have negative effects on investment. Finally, the effects on poverty and income distribution obtained through microsimulations are consistent with the results of the CGE experiments.
Keywords: Uruguay, labour market, general equilibrium model, regional integration,
efficiency wage, microsimulation, poverty
JEL Classification: D58, I32, F15, F16, J41
1
Introduction
In the nineties, with the signing of the MERCOSUR agreement, Uruguay deepened its
economic integration within the region and hastened the country’s trade liberalization
process. As a result, trade within MERCOSUR increased significantly but also increased
with the rest of the world. Trade openness led to resource reallocation from the
manufacturing sector towards services, profoundly affecting labour market structure. The
regional economic crisis that started with the Brazilian currency devaluation in 1999 led to a
four-year economic recession, worsening labour market and poverty indicators.
The purpose of this study is to find out how the labour market is affected by external shocks,
particularly those associated with the integration process or by changes in trade policies of
the bloc. It intends to estimate the effects on specialization, trade, employment and wages
stemming from those shocks or from changes in trade policies, taking into account the
imperfections and specific features of the labour market in different sectors of the Uruguayan
economy. It also identifies the impacts of these policy changes on poverty and income
distribution. Finally, the study evaluates policy options to lower the costs associated with this
process, directed to improve employment.
The study began with a review of the main characteristics of the Uruguayan labour market
and an estimation of wage differentials between sectors and labour categories in order to
obtain the stylised facts to be considered in the model. A CGE model was then built with the
purpose of running different scenarios of regional shocks, and trade and labour market
policies. Finally, microsimulations were run in order to evaluate the impact of these shocks
on poverty and income distribution.
In Section 2 a brief overview of the Uruguayan economy is presented and the main features
of the labour market are analysed, indicating the existence of imperfections that should be
taken into account for the specification of the CGE model used in the analysis.
Section 3 describes the main characteristics of the CGE model, the calibration and the
design of simulations carried out. It also presents the main aspects of the microsimulations
methodology that was adopted to analyse the impact on poverty and income distribution.
Section 4 presents the results obtained and, finally, Section 5 the main conclusions.
2
Economic overview
Recent economic performance
During the last 25 years Uruguay gradually adopted several reforms focused on the
liberalization and opening of real and financial flows in order to increase Uruguay’s ties with
the world economy, achieve macroeconomic stability and set the market as the main
mechanism for resource allocation. The process started in the mid-seventies, with great
transformations in the financial sector but only minor progress in terms of openness to trade.
By the end of the 70s, financial flows were completely liberalized, while trade reforms were
carried out more gradually. Starting from a maximum of 150 percent in 1980, by January
1993 the highest tariff was set at 20 percent.
The 1990s were dominated by the creation of the MERCOSUR, an imperfect customs union
among Argentina, Brazil, Paraguay and Uruguay. The creation of the MERCOSUR implied
the existence of free trade within the bloc and the adoption of a common external tariff
(CET), which was agreed in 1994 and enforced in 1995. The adopted CET varied from 0 to
20 percent, with an average tariff of 11 percent. However, many exceptions to its application
were accepted, and presently the four countries still apply different external tariffs to some
goods, mainly capital goods and computing and telecommunication goods. The full
enforcement of the CET by 2010 will mean that Argentina, Paraguay and Uruguay would
have to increase their tariffs on these goods, an unwelcome development since these
countries believe that such will hinder competitiveness in most sectors.
Since the creation of the MERCOSUR Uruguayan exports improved their access to a very
large market (the sum of Argentina and Brazil). Trade within MERCOSUR increased
significantly, and by 1998, 55 percent of Uruguay’s goods were destined for the bloc. This
was, in part, because of MERCOSUR, but this was also due to the loss of competitiveness
of Uruguayan exports to third countries. The latter was a result of an overvalued local
currency, the consequence of a policy of price stabilization based on the exchange rate.
Since similar stabilization policies were adopted in Brazil and Argentina, exports to those
countries remained competitive.
The situation changed dramatically when the Brazilian currency started to float in January
1999, affecting Uruguayan exports directly and indirectly (through the Brazilian impact on
3
Argentina). The share of Uruguayan goods exported to Brazil declined from more than a
third in 1998 to a little more than 20 percent in 2001. In 2002, the financial crisis in Argentina
also affected the Uruguayan economy. The reduction of Argentine income levels, the
restrictions on credit access, and the change in relative prices in that country had a negative
impact on Uruguay’s trade outflows. The total service exports (tourism basically) fell more
than 35 percent in the first quarter of 2002 compared to the same period in 2001 (that year,
80 percent of the tourists were Argentines). Exports of goods to Argentina dropped about 70
percent in the first semester of 2002 relative to the first semester of the previous year.
The Argentine crisis had relevant effects on financial activity as well. By 2001, the share of
deposit stock of non-residents from Argentina was high, but in February 2002 Uruguay
experienced an important capital flight due to the withdrawal of non-resident deposits. The
critical situation worsened by fraud in three of the main private Uruguayan banks. By August,
the deposit stock in the Uruguayan banking system had been reduced by 50 percent relative
to the beginning of the year, which forced the abandonment of the exchange rate system in
June 2002. A floating exchange rate was adopted, leading to a significant depreciation of the
local currency. The exchange rate accumulated a total 106 percent increase from December
2001 to December 2002.
Towards the end of July 2002, given the international reserves loss due to capital flight, the
Uruguayan government decided to call a “banking holiday” in order to make significant
changes in the banking system. The restructuring entailed the compulsory reprogramming of
term deposits in state-owned banks and the closure of four insolvent private banks (with a
very large market-share). After the restructuring, the branches of international banks
increased their importance in the banking system. This new financial situation implied severe
restrictions on bank credit access, which did not affect uniformly across sectors.
Although in the long run the Uruguayan economy shows a poor performance, e.g. the
average increase of GDP was 1.9 percent between 1970 and 1990, the 1990s were a period
of economic growth. Between 1990 and 1998 GDP increased by an annual rate higher than
4 percent (see Table 1). However, by the end of 1998 this process began to reverse itself
and after the Brazilian currency devaluation of January 1999 the country was in complete
recession. In 2003 economic recovery started in Uruguay, mainly driven by exports, which
grew 18 percent. Uruguayan exports had an 80 percent competitiveness gain in relation to
4
Brazil and other trade partners as a result of the depreciation of the Uruguayan currency.
Total GDP increased 2,2% in 2003 and 12,3% in 2004
Table 1 Main indicators
Year GDP a/ Annual inflation
a/
Fiscal balance
b/
Current account balance
b/
Imp. goods & serv. b/
Exp. goods & serv. b/
Gross capital
formation b/
Unempl. rate c/
1990 0,3 112,5 -3,0 2,0 18,10 23,53 12,20 8,51991 3,5 102,0 -1,8 0,7 17,86 20,69 15,13 8,91992 7,9 68,5 0,3 -0,8 19,63 20,45 15,38 9,01993 2,7 54,1 -1,7 -1,8 19,56 19,13 15,64 8,31994 7,3 44,7 -2,8 -2,3 20,38 19,77 15,87 9,21995 -1,4 42,2 -1,5 -1,3 19,10 19,00 15,41 10,31996 5,6 28,3 -1,4 -1,2 19,86 19,67 15,24 11,91997 5,0 19,8 -1,4 -1,1 20,54 20,55 15,22 11,41998 4,5 10,8 -0,9 -1,8 20,58 19,85 15,87 10,11999 -2,8 5,7 -4,0 -2,3 19,30 18,03 15,14 11,32000 -1,4 4,8 -4,0 -2,8 20,98 19,30 13,96 13,62001 -3,4 3,6 -4,3 -2,6 20,04 18,35 13,77 15,32002 -11,0 25,9 -4,2 3,1 20,01 21,97 11,52 16,92003 2,2 10,2 -3,2 -0,5 24,56 26,07 12,59 16,92004 12,3 7,6 -1,8 -0,8 27,94 29,65 13,29 13,1
Source: Elaborated with data from BCU and INE.
a/ Annual cumulative variation; b/ Percentage of GDP (current prices); c/ Urban areas
Recent trends in the Uruguayan labour market
In the nineties, the economic reforms carried out in Uruguay, along with increased trade
openness and the creation of the MERCOSUR led to a restructuring process that
determined changes in the composition of GDP as well as in the use of technology (Cassoni
and Fachola, 1997; Croce, Macedo and Triunfo, 2000; Tansini and Triunfo, 1998a; 1998b).
Between 1991 and 2002, the share of manufacturing employment was gradually reduced,
from 21 percent to 13 percent of total urban employment1. On the other hand, the share of
services, especially in retail, restaurants, hotels and financial services, increased: these
sectors, together with the construction sector, rose from 27.5 percent of total employment in
1986 to 39 percent in 2002. This, in turn, drastically affected the Uruguayan labor market,
displacing workers from some economic activities and changing the requirements of the
work force.
1 In 2001 the share of manufacturing employment was 17%. Between 2001 and 2002 methodology to measure industry product was modified by the INE, so the fall in 2002 might be overvalued.
5
The following facts thus characterized the evolution of the Uruguayan labor market in the
nineties: a) a generalized increase in labor productivity (output per worker); b) an increase in
the unemployment rate associated with the destruction of unskilled jobs; c) an increase in
wage dispersion, with a relative improvement of skilled wages; and d) an increase in
informality. These trends have deepened in the current decade.
Regarding the skill level of workers2, data analysis shows that the unemployment rate is
considerably lower for skilled workers, whereas unskilled workers show the highest
unemployment rates. The unemployment rate in Uruguay climbed from 8.8 percent in 1991
to 16.9 percent in 2002. Even prior to the severe economic crisis that affected Uruguay
between 1998 and 2002, the unemployment rate showed an increasing trend in a context of
economic growth. This evolution differed clearly according to the education level of the
labour force.
Another relevant change that occurred during the nineties was the reduction in public
employment as a result of the ongoing state reform process. Public employment share in
total employment fell from 24 percent in 1986 to 18 percent in 2002. However, public
employment for skilled workers rose slightly during the same period. As a consequence, this
structural change reinforced the effects of the changes observed in tradable sectors: greater
destruction of unskilled jobs (UNDP, 2001) 3.
It should be noted that the changes in productive structure affected not only the quantity but
also the quality of employment. Several studies suggest that precarious jobs, informality and
underemployment increased throughout the decade, especially for workers with low
education levels.4 The destruction of low skilled jobs that took place both in the public sector
and in the tradable sector drove unskilled workers towards employment in small productive
units or self – employment, thus leading to an increase in precariousness and informality
(UNDP, 2001; Bucheli, 2005). In this context, informality became one of the most important
imperfections in the Uruguayan labour market, affecting more than one third of employed
workers during this period.
2 This study considers that a worker is skilled if he has at least 12 years of formal education. 3 In 1997-99, 22,400 unskilled and 5,600 skilled public jobs were destroyed. During the same period 3,600 new skilled jobs were created. The outcome was the destruction of 24,400 public jobs. 4 Precarious workers are those private dependent workers who are not covered by social security, have an unstable job, or receive no remuneration for their work. Informal workers are those not covered by social security. Underemployment comprises workers who work less than 40 hours a week but would be willing to work additional hours.
6
Methodology
Labour market specification
As cited in the previous section, the Uruguayan labour market presents serious problems of
unemployment and informality. Therefore, we considered that these imperfections had to be
captured in the model. However, we needed to focus on one type of imperfection. Since one
of the distinguishing features of the Uruguayan labour market is the existence of a persistent
wage differential between formal and informal jobs, a dual market labour approach was
adopted for the study.
The theory of dual labour markets is based on a two-tier regime where a primary and a
secondary sector coexist with unemployment. Workers in the upper tier (primary sector)
enjoy higher wages and fringe benefits; also, stability, union protection and labour regulation
enforcement are more likely in this sector. Meanwhile, in the low wage (secondary) sector,
the labour market clears and workers in this sector are not able to underbid those in the
primary sector. A rationing of jobs in the primary sector explains the existence of queues and
the persistence of unemployment. On the other hand, the secondary sector provides
flexibility to the economy, adjusting its size to fluctuations in the business cycle.
The efficiency wage model provides a microeconomic foundation for these features.
Different versions propose a persistent wage gap and a negative relationship between the
rate of unemployment and the upper tier wage (Saint Paul, 1996). In the Shapiro-Stiglitz
version firms are interested in paying wages above the expected rates because of cost
monitoring reasons (Shapiro and Stiglitz, 1984). The model assumes some inability of
employers to observe workers effort. This allows the worker to choose how much effort he
wants to make but if he shirks, he faces the probability of being fired. There is a critical wage
above which the worker will not shirk. As far as firms are concerned, they compete by
offering wage packages that take into account the minimum wage required to induce
workers’ effort.
In equilibrium, if wages are very high, workers will value their jobs not only by the high wage
itself but also by the low level of employment (due to low labour demand at high wages).
Among others reasons, the critical wage for non shirking will be greater; i) the lower the
probability of being caught shirking; ii) the higher the expected utility of being unemployed
7
(e.g. the most generous the unemployment subsidy program) and iii) the higher the flows out
of unemployment.
The same results arise in versions based on “recruit, retain and motivate” reasons. On one
hand, a high wage eases vacancies filling, reduces the quit rate and motivates effort. On the
other hand, high unemployment affects the likelihood of finding a new job if dismissed (thus
affecting effort) and the ease of voluntary turnover (quit).
In the CGE model we assumed that an efficiency wage model might explain the wage gap
between formal and informal workers. In order to estimate the wage gap, the study used the
Continuous Household Survey (CHS) collected by the National Statistics Institute (INE) in
2003. We restricted the sample to wage earners and self-employed workers between 18 and
59 years of age. Informal workers are defined in this paper as those who do not contribute to
the social security system.
Different estimations of the wage gap were used in the simulation. First, we used a very
simple econometric model: we regressed by ordinary least squares (OLQ) the log hourly
wage on individual and labour characteristics, including a dummy variable that identified if
the worker was formally employed.5 The estimated dummy’ parameter (G1) is a measure of
the wage gap. Then we used the usual way of decomposing wage differences, proposed by
Oaxaca (1973) and Blinder (1973). The study estimated an earning equation for the formal
workers and another one for informal workers. The difference of the characteristic’s rewards
--weighted by the mean of the formal workers -- is interpreted as the mean wage gap (G2).
Analogously, the study estimated the difference between coefficients but weighted by the
average characteristics of informal workers (G3)
This estimation ignores the endogeneity of the selection decision being formal or informal.
We expect unobservable, individual characteristics to be correlated with being formal or
informal (i.e. people with easy access to informal networks or to informal benefits could have
more potential gains than being informal). To deal with this problem we estimated a
switching regression model and used it to calculate the gap between the predicted wage of
an average informal worker and the wage he would have had in the formal sector (G4).
5 We controlled personal characteristics (age, education, gender, marital status, geographical region), the type of occupation (public servants, size of the establishment of the private wage earners, self-
8
The results of the estimations are shown in Appendix 1. We report the different estimations
of the wage gap in Table 1. The four alternative estimations suggest that formal workers are
highly remunerated.
Table 2 Estimated mean difference in earnings between formal and informal workers (log Wf-log Wi) Raw gap 0,85 Estimated gap G1 0,59 G2 0,65 G3 0,60 G4 0,52
A CGE model that captures these conclusions is presented in the next section.
The CGE model
A Computable General Equilibrium (CGE) model was used to analyse the effects of several
external shocks and some specific policies on the Uruguayan labour market,. It is based on
the model by Laens and Terra (2000), with several changes regarding labour market
behaviour, export demand and institutional design.
The structure of the core CGE model is quite conventional in terms of the analysis of trade-
related issues, but an alternative specification is made regarding the labour market. We used
two different versions of the model for the simulations: an efficiency wage6 model and a
competitive labour market model.
The main features of this model are as follows:
It is a multi-sector model with 23 sectors, including two special sectors. One of the
two special sectors gathers all the activities (mainly, public services and the financial
sector) where employment and wages are fixed, because institutional arrangements
and/or trade unions are a deterrent to workers’ dismissal or to wage reductions. By
law, public employment is fixed: no new public employers are hired, and the existing
employed who own some property and self-employed who do not) and other labour characteristics (part-time and industry). 6 Following Thierfelder, K.E., C.R. Shiells (1997) and Annabi, N. (2003)
9
ones cannot be fired. Although trade unions could have been introduced in the
model, our intention was to focus on labour market duality between formal and
informal workers. Trade union modelling might be included in a future specification of
the model.
The other special sector is an informal sector that produces one type of good
destined to domestic final consumption.
We assume that Uruguay has three trading partners (Argentina, Brazil and the rest of
the world). The Uruguayan economy is explicitly modeled while in the case of the
other trading partners only the supply of imports and the demand for exports are
endogenous.
Perfect competition is assumed in all sectors. However, goods are not homogenous,
as they are differentiated by geographic origin.
We assume that there are ten representative households which represent different
income levels (by deciles of the income distribution).
Government collects tariffs and taxes. Government revenue is used to buy goods
and services and to make transfers to households. We assume that government has
fixed consumption of goods and services (in physical units) and the transfers to
households are updated by the change in the average wage7. Government savings is
obtained as a residual.
On the production side, the study uses a nested production function. At the top level,
firms combine intermediate inputs with value added following a Cobb-Douglas
function. Value added is obtained with a constant elasticity of substitution (CES)
function that combines capital and composite labour. Then, composite labour is
obtained by combining skilled and unskilled labour with a CES. In the informal sector,
value added is only composed by unskilled labour.
7 In 2001 social security transfers represented nearly 83 percent of total government transfers to households. In 1989 a constitutional reform measure established that social security benefits are adjusted to the evolution of Average Wage Index.
10
Goods are imperfect substitutes in consumption (Armington). The small country
assumption is made for imports, so the country faces a perfectly elastic supply curve
in the external markets. However, it is assumed that the country faces a downward
sloping demand curve for exports (quasi small open economy)8. Export demand is a
function of relative prices and real income in the trade partners, which are considered
exogenous.
Total demand for each sector is composed by domestic demand (intermediate and
final) plus exports to each of the trading partners.
Trade balance is fixed, so that imports and exports of goods and services maintain
the difference existing in the benchmark data. The equilibrium in the model is defined
by the simultaneous equilibrium in goods and factor markets and in the external
sector.
There are three factors of production: capital, skilled, and unskilled labour (the labour
market is segmented by qualifications). The supply of each factor is fixed and there is
no international mobility. Skilled labour can only be employed in the formal sectors,
while unskilled labour can also be employed in the informal sector.
In the model with efficiency wages, this behaviour is applied to all formal activities,
except for those in the fixed employment sector, which we named APUBLIC,
because it is mainly composed of public activities. Unemployment is fixed, so when
unskilled workers are fired from the efficiency wage sectors, they go to the informal
sector where they receive a lower wage. The specification of efficiency wage
behaviour follows Thierfelder and Shiells (1997).
The model was run using GAMS (General Algebraic Modeling System).
Calibration of the CGE model
The model was calibrated using a Social Accounting Matrix (SAM) with data for the year
2000. It was taken from Barrenechea, Katz and Pastori (2004). Originally, the SAM included
8 Following Cox’s specification (1994).
11
30 different activities and 36 different commodities. Even though this disaggregation was
quite appropriate for this study, some adjustments had to be made.
Specifically, it was necessary to show the differences in labour, according to the qualification
of workers and their status of formality or informality. Therefore, labour was separated into
skilled and unskilled labor. Among skilled workers, informality is not easily available, so it
was assumed that skilled labour is always formal. Information about qualifications and
formality of workers was taken from the 2003 Continuous Household Survey (CHS) collected
by the National Statistics Institute (INE). Workers with twelve or more years of formal
education were considered skilled workers.
In order to study the labour market, it was also necessary to distinguish between private and
public activities, because there are rigidities concerning both wages and employment in the
public sector. Some activities, which are carried out by public and private agents (for
example, education or electricity supply), needed to be decomposed. Therefore, a new
activity was created that included all activities carried out by the public sector9. This sector
combines skilled and unskilled labour, such as those found in the private sector, but public
employment is considered fixed. To separate public from private activities, information was
also taken from the 2003 CHS.
In addition, government final consumption was disaggregated in the new matrix. In the
original SAM, government final consumption expenditure was included in a miscellaneous
sector called “other services”. Final consumption expenditure of government was estimated
from National Accounts. Then, final consumption expenditure was disaggregated according
to the information provided by the 1995 SAM (Lorenzo, Osimani and Caputti, 1999; Laens
and Terra, 2000).
The rest of the world needed to be disaggregated as well. Argentina and Brazil were
separated from the rest of the world, creating three foreign agents. In this case, data was
taken from National Accounts and trade statistics from the Central Bank (BCU).
Finally, an informal sector was created besides those originally considered in the SAM. It
was assumed that the informal sector produces a composite good of all the activities in
9 That is: Electricity and Water Supply, Petroleum Refinery, Communications, Postal Service, Financial Services and Educational Services.
12
which informal labour was identified. This “informal good” is produced entirely for final
consumption of households. It was assumed that value added of the informal sector includes
only wages. The total amount of informal sector wages was estimated with data from the
CHS. As a result, the informal sector includes activities such as agriculture and other primary
activities, construction, retail, and textiles and clothing, which have an important component
of informality.
The microsimulations methodology
The CGE model provides some insights about the poverty effects of the shocks and policies
that were simulated. However, the combination of these results with a microsimulation
methodology provides more precise information about poverty and income distribution by
tracking the economy-wide changes at the household level. Several approaches have been
developed with this purpose, as shown by Bourguignon, Pereira de Silva and Stern (2002).
The microsimulations are based on household data but there is no need to reconcile this
data with the SAM because the procedure only needs information about changes in wages,
employment and unemployment. The method assumes that changes in the labour market
can be replicated by a random selection procedure, which imposes counterfactual changes
in labour market parameters calculated for the benchmark year. This approach follows Paes
de Barros and Leite (1998), Paes de Barros (1999), Frenkel and González (2000), Ganuza
et al (2002) and Ganuza et al (2004). It was applied for the case of Uruguay by Bucheli et al
(2002) and by Laens and Perera (2004). The SPSS program used in this paper is the same
one used in the latter work.
The rationale for using microsimulations is that a CGE model captures only partial
distribution of income between families, therefore making it difficult to see the real impact of
shocks or public policies on income distribution and poverty. A crucial assumption adopted in
this methodology is that a person’s position in the labour market is the main determinant of
his income and poverty status.
The procedure takes CGE results as inputs. Labour market structure is considered as a
function of six parameters: participation rate, unemployment rate, wage structure, overall
average wage, worker’s education level and structure of employment (sector of activity and
occupation category). In this study, the participation rate is fixed, so it is not considered for
13
the microsimulations. In turn, sector of activity is defined in terms of formal or informal
activity.
Once the changes in the labour market parameters are obtained from the CGE results, the
microsimulation methodology is applied. The procedure uses random numbers to simulate
the changes in the labour market structure that are consistent with the parameters
introduced. On average, the effect of the random changes will reflect the impact of the new
(simulated) parameters in the labour market. The microsimulations are repeated a large
number of times using Monte Carlo numbers to allow for the determination of confidence
intervals for the poverty and income distribution indicators. In each simulation, changes in
poverty and income distribution are measured through the percentage of population under
the poverty line, the poverty gap, the Gini coefficient and the Theil coefficient. Data from
CHS for the year 2000 was used.
For each scenario, several changes in labour market structure were simulated, first
separately, then sequentially. The idea behind establishing a sequence is that changes in
labour parameters follow some order, which is not neutral. The commonly accepted
sequence is the following: first the person decides whether to participate or not in the labour
force; then the market decides whether he or she will be employed or not; then the person
decides whether to work in the formal or formal sector and this determines a certain wage
level and, in the aggregate, the average wage. Finally, labour market structure by education
level is defined. This sequence was applied in the three models considered. As
unemployment is fixed in the model, the corresponding rate remains unchanged. The
analysis was made taking the whole sequence into account.
Simulation design and results
Simulation design
In Section 2 we pointed out that increasing trade openness and the integration of the
Uruguayan economy to the MERCOSUR augmented the country’s vulnerability to external
shocks, particularly those originating in Argentina or Brazil. With that idea in mind, the study
carried out some simulations in order to show how and why some of the forces at work
during the 2002 crisis affected the Uruguayan labour market.
14
Again as explained in Section 2 the crisis had many components: recession in Argentina and
Brazil, changes in relative prices that affected Uruguayan exports to those countries, credit
constraints, financial turmoil, external debt growth, capital flight, etc. Unfortunately, it is
impossible to evaluate with our CGE the specific weight of each of these factors in the
genesis and the deepening of the crisis, particularly because it had a very significant
financial component, which cannot be tracked by this model.
Nevertheless, we chose to simulate two relevant components of the 2002 crisis: the change
in relative prices vis-à-vis the main trade partners (due to devaluations in those countries)
and the foreign savings constraint. In order to assess the effects of the change in relative
prices that occurred when Argentina abandoned the currency board regime, we simulated a
40 percent decline in domestic prices nominated in dollars in Argentina and a 7 percent
decrease in the price of imports from that origin (ARGRP scenario), which was what really
happened in Argentina between 2000 and 200210. In order to compare the effects of the
shocks originating in one or the other MERCOSUR partner, we simulated an identical
change in prices in Brazil (BRARP).
The third simulation was a restriction in foreign savings. In 2000 the Uruguayan current
account was running a deficit, which was financed by capital inflow from the rest of the
world. In 2002 the situation was reversed and no capital inflow was available, so a severe
adjustment was needed to obtain a current account surplus. Therefore, in this simulation we
fixed the current account balance (EXTSAV) at zero.
As cited in Section 2, the MERCOSUR is an imperfect customs union because the common
external tariff has not been fully enforced across the four countries. Its full enforcement was
simulated in order to assess the effects that it might have on the Uruguayan labour market,
especially because the rise in capital goods tariffs might have a negative effect due to a
competitiveness loss (CET).
Finally, we simulated a specific labour market policy. Assuming that a reduction in the
relative cost of labour might improve employment, the study simulated a 10 percent direct
subsidy on formal employment of unskilled labour (DIRTAX).
15
Table 3 summarizes the five experiments and shows how variables or exogenous
parameters are affected. The complete model equations are presented in Appendix 2.
Table 3
Simulation scenarios
The results of these five simulations with the CGE model are presented in Tables 4
(Variation of main macroeconomic variables) and 5 (Effect on labour market variables).
Simulations of regional shocks and results
These experiments show the vulnerability of the Uruguayan economy to regional shocks,
which have increased due to geography and the deepening of the integration process with
the MERCOSUR countries.
A change of relative prices in any of the MERCOSUR partners generates a GDP decline in
Uruguay, a reduction of exports and imports, and a decrease in investment. The reduction of
both exports and imports is due to our choice of model closure, which fixes current account
balance. When export demand falls as a consequence of the relative price change with the
trading partner, imports fall as well, and adjustment is done through the exchange rate, with
a devaluation of local currency.
10 Data was taken from Indec- National Institute of Statistics and Censuses of Argentina.
Simulation scenario Variable or exogenous parameters Variation (%)ARGRP Domestic price index (DPARG i) -40%
Exports price from Argentina (PWARG i) -7%BRARP Domestic price index (DPBR i) -40%
Exports price from Brazil (DPBR i) -7%EXTSAV Current account balance (B) -100% *CET Common external tariff (t)DIRTAX Labour taxes (trab) -10%
* In benchmark the current account balance was 4% of GDP
16
Table 4
The macroeconomic impact of the same change in relative prices is more pronounced when
it happens in Argentina than when it happens in Brazil. This could be explained by the
relative importance of exports to each country in the benchmark: 24 percent of total exports
were destined for Argentina, 17 percent for Brazil and 59 percent for the rest of the world. In
turn, the share of imports from those origins was 26 percent, 18 percent and 56 percent,
respectively.
This result should be taken with caution because it is not necessarily true that a shock
coming from Brazil will always have lower effects on the Uruguayan economy than a shock
from Argentina. This result is highly dependent on the prevailing macroeconomic conditions,
as the region has been affected by severe instabilities that have significantly changed the
trade composition by origin or destination. As long as Brazil increases its relative importance
as trade partner for Uruguay, the impact of a relative price change in that country could
increase substantially.
Relative prices change with Argentina
Relative prices change with
Brazil
External Savings
Restriction
Common External Tariff
Subsidy to Unskilled
labour
Absortion* -0,38 -0,24 -4,42 -0,23 -0,03Household Consumption* -0,14 -0,12 -0,66 -0,18 1,65Investment* -2,24 -1,19 -31,32 -0,77 -10,00Exports* -7,26 -2,79 9,97 -3,18 -0,44Imports* -5,95 -2,04 -11,82 -2,54 -0,36Real GDP -0,48 -0,33 -0,54 -0,29 -0,03Real Exchange Rate 4,19 2,15 1,45 -0,36 -0,81Export Price -0,12 -0,09 0,00 0,00 0,00Import Price -2,96 -2,44 0,00 0,00 0,00Consumer Price 0,05 0,00 -0,25 0,09 -0,21
Absortion* -1,13 -0,30 -4,59 -0,22 0,20Household Consumption* -0,28 -0,23 -0,58 -0,17 1,85Investment* -7,38 -1,03 -33,13 -0,75 -9,39Exports* -8,99 -4,62 10,25 -2,78 -0,34Imports* -8,22 -4,27 -11,91 -2,14 -0,26Real GDP -1,11 -0,27 -0,64 -0,29 0,19Real Exchange Rate 4,37 2,57 1,57 -0,45 -0,60Export Price -0,40 -0,29 0,00 0,00 0,00Import Price -2,67 -2,16 0,00 0,00 0,00Consumer Price 0,04 0,03 -0,14 0,09 -0,17
Perfect competition model
Efficiency wage model
Macroeconomic variables for each simulationPercent Variation
17
The impact of a relative price change in Argentina is higher when efficiency wages and the
existence of an informal sector are assumed. In this case, real GDP falls by 1.1 percent,
while it decreases 0.38 percent when the neoclassical assumptions are adopted. The
variation of Argentine relative prices generates a very significant reduction of investment in
Uruguay, which would reach 7.4 percent in the efficiency wage model and 2.2 percent in the
perfect competition model. Investment declines because government savings decline (as
government revenue is lower) along with household savings.
Table 5
On the other hand, a change of relative prices in Brazil has greater impact on the Uruguayan
GDP when the perfect competition model is used. This could be explained by the factor
intensity of goods traded, which is quite different from one country to the other. Trade flows
with Argentina are more intensive in skilled labour than trade flows with Brazil (see tables 6
and 7). Therefore, a competitiveness loss with Argentina generates a reallocation of
resources towards industries that make intensive use of unskilled labour and capital (see
table 8).
Relative prices change
with Argentina
Relative prices change
with Brazil
External Savings
Restriction
Common External
Tariff
Subsidy to Unskilled
labour
Informal Emp -0,48 0,21 0,48 -0,01 -4,97Unskilled Emp 0,10 -0,04 -0,10 0,00 1,02Unskilled Wage 0,34 -0,33 -1,14 -0,17 6,98Skilled Wage -0,95 0,20 -0,33 -0,23 0,07
Informal Emp -0,28 0,19 0,37 0,02 -4,36Unskilled Emp 0,12 -0,08 -0,16 -0,01 1,86Wage Differential 0,16 -0,11 -0,21 -0,01 2,75Unskilled Wage 0,00 -0,42 -0,94 -0,19 6,50Skilled Wage -1,00 0,09 -0,23 -0,23 -2,76
Efficiency wage model
Labour Market Variables for each simulationPercent Variation
Perfect competition model
18
Table 6
Argentina Brazil Rest of the worldSkilled labour / Unskilled labour 0.68 0.30 0.37Capital / Unskilled labour 0.94 1.47 1.82
Factor Intensity of export by destination
Table 7
ARG BRA RM ARG BRA RM ARG BRA RM
Agriculture & agroindustries 0.08 0.67 1 338 964 11.9 55.9 66.1 12.1 14.7 9.4Other manuf. goods 0.37 1.33 -411 -330 -1289 25.8 34.7 17.6 63.7 73.1 70.6Services 1.33 1.06 434 -24 -162 62.3 9.4 16.4 24.2 12.1 20.0Total 1.00 1.00 23 -16 -487 100 100 100 100 100 100
Specialization and Factor Intensity by sector
% of Exports Trade Balance
(millions of U$S) % of ImportsSkill Labour/
Non Skill Labour
Capital/ Non Skill Labour
The study assumes that the skilled labour market is perfectly competitive while the unskilled
labour segment is subject to efficiency wages, so that an increase in demand for unskilled
labour and a reallocation of resources to those sectors make the results differ more than in
the case when the reallocation of resources operates in the opposite direction.
In fact, when there are reasons for paying an efficiency wage, an inefficient resource
allocation takes place. The production possibilities frontier shifts to the left when
specialization becomes biased towards the production of goods intensive in unskilled labour.
Therefore, the larger the specialization in goods intensive in unskilled labour, the greater the
inefficiency generated by the existence of efficiency wages and the greater the difference in
GDP in relation to an economy where the labour market is perfectly competitive.
19
Table 8
In order to simplify the problem, we can gather production in two big sectors according to
their intensity in skilled or unskilled labour. The following graph illustrates the argument:
The curve PPF1 is the production possibilities frontier when the labour market is perfectly
competitive, while PPF2 shows the production possibilities frontier when there are efficiency
wages in the unskilled labour market segment. Production possibilities are reduced more as
production gets more specialized in goods that intensively make use of unskilled labour. P0
PPF1
PPF2
Goods intensive in
unskilled labour
Goods intensive
in skilled labour
P0 P’0
P1 P’1
AGRI MANUF SERV INFORMALShare of sector in
total output 15,4 16,4 63,8 4,4
Perfect Comp 5,1 2,4 -0,8 -0,3Efficiency Wage 6,1 1,5 -1,9 -0,2
Perfect Comp -2,9 -1,6 0,2 0,1Efficiency Wage -3,6 -1,4 0,3 0
Perfect Comp 2,1 0,8 -1,5 0,5Efficiency Wage 2,8 0,9 -1,8 0,3
Perfect Comp -0,6 1,2 -0,1 -0,1Efficiency Wage -0,8 1 -0,2 0
Perfect Comp 2 0,3 -0,4 -2,3Efficiency Wage 1,1 0,2 0 -1,9
External Savings Restriction
CET
Subsidy to Unskilled labour
Output shares and variation by sectorPercentages
Argentina RP
Brazil RP
20
and P’0 show the best production combinations under perfect competition and under
efficiency wages for the initial relative prices. The graph shows that as relative prices
change, favoring an increase in the production of goods intensive in unskilled labour, the
production combinations shift to P1 and P’1, respectively. It can be observed that P1’ is more
distant from P1 than P0’ is from P0, due to the bias in the production possibilities frontier. This
is because when employment increases in the efficiency wages sector, there is an efficiency
loss due to an increase in the wage differential.
Table 4 shows that the Argentine change in relative prices generates a very significant
reduction in Uruguayan exports, which brings about an increase in specialization in goods
intensive in unskilled labour (see table 7). In 2000, 62 percent of total exports to Argentina
were services (tourism, financial services, transportation, etc.), many of them intensive in the
use of skilled labour (especially, financial services).
Table 5 shows the corresponding effects of these shocks on the labour market. In perfect
competition, a change in relative prices with Argentina generates the opposite effect than the
same change in Brazil: labour demand increases and so does the wage of unskilled labour,
relative to skilled labour wage. A similar occurrence can be found in the version of the model
with efficiency wages.
The experiment that assumes an external savings restriction, due to the uncertainty
prevailing in the region, generates a very significant decline in imports and investment, while
there is an increase in exports. The effects on informal employment and wages are similar to
those obtained in the case of a Brazilian change in relative prices, but their size is bigger. In
this case, there is also a reallocation of resources towards the traditional exporting sectors,
which are intensive in unskilled labour. Sectors like meat packing, dairy products, rice and
other typical exporters, increase their unskilled labour demand by more than five percentage
points. However, the reduction of investment brings about a 25 percent decrease in unskilled
labour demand in construction as 75 percent of this sector’s output is destined for
investment. This leads to a reduction in the service sector, but this reduction is concentrated
to service sector activities that are intensive in unskilled labour. Therefore, unskilled labour
demand falls, increasing informality. In addition, the external savings decline generates a fall
in payments to all factors (see table 5).
21
Simulation of MERCOSUR deepening and results
Simulating the full enforcement of the MERCOSUR CET implies an increase in protection in
the Uruguayan domestic market, but its global impact is scarce (minimal?) (see table 4).
Absorption, household consumption, trade and GDP fall, and this happens in the two
versions of the model. There is a reallocation of resources towards the manufacturing sector
(chemicals and other import substituting industries), leading to a more intensive use of
capital. Anyway, the change in the production structure is slight. (see table 8). The increase
in protection brings about an anti-export bias, so agriculture falls. Within services, the
sectors that grow are commerce and transportation, but health services, hotels and gas
distribution fall and so does other services.
In the labour market a wage decrease is observed, mainly for skilled workers. In the
efficiency wage model, there is an increase in informal employment (see table 5). Therefore,
the CET approved by the MERCOSUR would not have a positive effect on employment in
Uruguay: it would protect workers in the manufacturing sector (where employment
increases) but would harm global employment.
Impact of employment policies
We tried to analyse the impact of some policies that could compensate for the negative
effects on unskilled labour wages and informal workers, which were found in the previous
simulations. To these ends, a 10 percent subsidy was simulated in the case of formal
employment of unskilled workers (DIRTAX). The rationale for this type of policy stems from
the existence of efficiency wages, which lead to lower employment of unskilled workers.
This policy would have a low impact on absorption and trade, and would increase household
consumption, but investment would fall (see table 4). Even though global income increases,
savings do not increase in the same proportion because this policy favors lower income
households: their income increases exponentially, but these households have lower
propensity to save. On the other hand, the policy has a strong fiscal impact, as government
expenditure and deficit increase. This explains the investment decline. Table 9 shows the
evolution of income for every agent. In the poorest households income increases by 3
22
percent, while in the richest households, it only increases 0.5 percent. In turn, net
government income (revenue minus the subsidy cost) falls by almost 6 percent.
Table 9
In the perfect competition model, this policy has a negative effect on GDP due to its negative
effects on efficiency and resource allocation, but it has a positive effect on GDP in the model
with efficiency wages, because this policy tackles the core of the market imperfection: the
demand for unskilled workers is below the optimum because there is a cost associated to
monitoring, hiring and training such workers.
In the labour market, a very significant increase in unskilled labour demand is observed,
which is translated into higher employment of unskilled workers in the formal sector and a
rise in their wage (see table 5). In perfect competition, the wage of unskilled workers rises by
7 percent, while this rise is at 9.4 percent in the efficiency wage model. This is consistent
with the informality decline of -2 percent in the efficiency wage case. In addition, these
changes increase the relative wage of unskilled workers.
Consequently, even though this type of policy seems appropriate in increasing efficiency and
improving income distribution, when the efficiency wage hypothesis is valid, it may have
perverse long run effects. This is so because investment falls, and there is also a
disincentive to human capital accumulation. Both aspects might hinder economic growth in
the long run.
Perfect Competition Efficiency Wage
Household average 1,5 1,7First decile 2,4 3Second decile 2,6 3,1Third decile 3,2 3,6Fourth decile 2,6 2,9Fifth decile 2,3 2,6Sixth decile 2,3 2,6Seventh decile 2,4 2,6Eight decile 2,1 2,3Nineth decile 1,2 1,3Tenth decile 0,4 0,5Government -5,9 -5,7
Income variation as a result of subsidy on unskilled labourPercentages
23
Impact on income distribution and poverty
In order to analyse the impact on poverty and income distribution of the shocks simulated
with the CGE, we ran microsimulations for two cases: the external savings restriction
(EXTSAV) and the subsidy to formal employment of unskilled labour (DIRTAX). In both
cases the microsimulations were run based on the CGE results obtained from the two
different versions of the model. We chose these two cases because these had the greatest
impact on employment, informality and wages.
For each microsimulation, changes in poverty are measured by two indicators: 1) the
percentage of people under the poverty line; and 2) the poverty gap that shows the average
distance between their income and the poverty line. Income distribution is measured with
two well-known indicators: the Gini coefficient and the Theil coefficient. Table 10 shows the
results obtained from these microsimulations. As can be observed, all the results are
significant with a 95 percent confidence interval.
Table 10
The restriction on external savings increases the share of the population below the poverty
line and the inequality in income distribution, whereas a subsidy on unskilled labour
employment in the formal sector has the opposite result. This is consistent with the changes
in relative wages between skilled and unskilled labour found in the CGE results.
In the efficiency wage model, a reduction in external savings leads to an increase in poverty:
the population below the poverty line increases by 1.1 percent. In addition, income
distribution deteriorates, as the Gini coefficient increases by 0.2 percent and the Theil
Base year values (%)
External savings restriction
Subsidy on unskilled
employment
External savings restriction
Subsidy on unskilled
employmentPopulation below poverty line (P0) 17,8 1,9 -8,3 1,1 -7,7Poverty gap (P1) 5,6 1 -9,5 1,3 -7,7Gini coefficient 44,2 0,1 -1,4 0,2 -1,9Theil coefficient 35,5 0,3 -2,8 0,7 -3,9
*All results are significant with a 95% confidence interval
Microsimulation results*
Perfect competition model Efficiency wage modelPercentage variations
24
coefficient by 0.7 percent. The results obtained with the perfect competition model are very
similar.
The microsimulations based on the CGE results for the subsidy on formal employment of
unskilled labour show a positive impact on poverty and income distribution. The population
below the poverty line declines - 8.3 percent in the perfect competition model and - 7.7
percent in the case of the efficiency wage model. Income distribution also improves, as the
Gini coefficient is reduced by -1.4 and -1.9 percent, respectively. This might be explained by
the significant rise of unskilled wage when this type of policy is implemented: in the efficiency
wage model, unskilled wage rises 6.50 percent while the wage differential between formal
and informal unskilled workers rises 2.75 percent.
Conclusions
The analysis of the Uruguayan labour market clearly shows the existence of wage
differentials between sectors and labour categories. These differentials are wider between
workers employed in the formal and in the informal sector, and between skilled and unskilled
labour. These characteristics of the Uruguayan labour market indicate the need to
incorporate labour market imperfections in the analysis of external shocks and trade policies
using a CGE model.
Minimum wage is not effective in Uruguay and labour unions are not strong enough to
explain those differentials, except in a few activities, and mainly covering the public sector.
Therefore, based on this evidence, the study assumed the existence of efficiency wage
behaviour in the private formal sector.
In this context, we constructed a CGE model in which we distinguished for kinds of workers.
First, we made a distinction between skilled and unskilled workers. Second, we noted that
there is a group of workers in a fixed employment sector, mainly the public sector. Then, as
informality is not important for skilled workers, we considered that duality only affects
unskilled workers. When unskilled workers are fired from the efficiency wage sectors, they
go to the informal sector where they receive a lower wage.
Different simulations were carried out with two versions of the CGE model: perfect
competition and efficiency wage. In the second model it was assumed that the labour market
25
segment for skilled labour operates without distortions, while unskilled labour behaves in an
efficiency wage mode. This assumption is reasonable, as both unemployment and
informality are low for skilled labour. The perfect competition model was run as a reference.
In the efficiency wage model, an extreme assumption was adopted concerning the
displacement of unskilled workers. It was assumed that all displaced unskilled workers went
to the informal sector. In fact, some of them remain unemployed.
One clear conclusion from the simulations carried out in this study is that the MERCOSUR
economies deeply affect the Uruguayan economy through changes in relative prices. The
study shows that the same shocks on relative prices are more important for Uruguay when
they originate in Argentina than when they occur in Brazil. However, this result should be
taken cautiously because it is highly dependent on the composition of trade with each of
those partners. In the benchmark year trade of goods and services was more important with
Argentina, which explains the greater impact of shocks from that origin.
Similarly, a restriction on external savings as a consequence of the instability in the region
has significant effects on the Uruguayan labour market. On the contrary, the full enforcement
of the common external tariff approved by the MERCOSUR does not have an impact of
relevance.
The first four simulations show the impact that macroeconomic instability in the region can
have on the Uruguayan economy. Both the effect of changes in relative prices with Argentina
and Brazil and an external savings restriction are significantly larger than a tariff change. The
implementation of policies that tend to reduce the region’s share in total trade, such as the
reduction of the CET, or free trade agreements with third markets (FTTA, EU-MERCOSUR
agreement) is therefore important for Uruguay. At the same time, one main objective of
Uruguayan macroeconomic policy should be to avoid significant changes in relative prices
with its main trading partners.
The consideration of labour market imperfections is particularly important in cases where the
simulations lead to a reallocation of resources towards sectors that use unskilled labour
intensively. In this case, the increase in the wage premium implies an efficiency loss, which
is larger in areas where the economy is more specialized, such as sectors intensive in
unskilled labour.
26
The simulation of a subsidy on formal employment of unskilled workers shows that despite
the increase in the wage premium, there is an increase in GDP due to the efficiency gain
derived from the decline of informal employment or unemployment. The introduction of a
subsidy stimulates demand for unskilled labour, thus compensating the demand reduction
caused by the inefficiency derived from the wage premium. Even though this policy leads to
an improvement in employment and income distribution, it generates a decline in investment
and a disincentive to human capital accumulation, which could be harmful for growth in the
long run.
However, this kind of policy could still be implemented but should be more focused on
specific workers, and with a lower tax rate. This way, adverse macroeconomic effects in the
long run could be avoided, and informal, low productivity employment could be reduced.
With this in mind, a more disaggregated CGE model can contribute in evaluating the impact
of more focused policies in the future.
Finally, the effects on poverty and income distribution obtained through microsimulations are
consistent with the results of the CGE experiments, that a restriction in foreign savings has a
negative effect on both. On the contrary, a policy that introduces a subsidy on formal
employment of unskilled labour reduces the percentage of population under the poverty line
and improves income distribution.
The study results show the importance of taking into account the existence of imperfections
in the labour market. The effects of external shocks, as well as the impact of some policies
are clearly different in the presence of these imperfections. This fact emphasizes the need to
make an appropriate diagnosis of the labour market when modeling the economy of a
particular country.
References
Annabi, N. (2003): “Modeling labor markets in CGE models. Endogenous labor supply,
unions and efficiency wages”, Québec, Université Laval (PEP).
Barrenechea, P., G. Katz, H. Pastori (2004): “Construcción de una matriz de contabilidad
social para Uruguay para el año 2000”, Working Paper 20/04, Department of
Economics, School of Social Sciences, Universidad de la República, Uruguay.
27
Blinder, A. (1973), “Wage discrimination: reduced form and structural estimates”, Journal of
Human Resources, 8(4).
Bourguignon, F., L. Pereira da Silva, N. Stern (2002): “Evaluating the poverty impact of
economic policies: Some analytical challenges”, Washington D.C., The World Bank
(Mimeo).
Bucheli, M. (2005): “Uruguay: La cobertura de la seguridad social en el empleo, 1991 –
2002”, in Bertranou, F. (ed.): Protección Social y Mercado Laboral, ILO, Santiago de
Chile.
Bucheli, M., R. Diez de Medina, C. Mendive (2002): “Uruguay: Equidad y pobreza ante la
apertura comercial de los noventa. Un enfoque a través de microsimulaciones.”, in R.
Vos, L. Taylor and R. Paes de Barros (eds.) Economic Liberalisation, Distribution and
Poverty: Latin America in the 1990s, Cheltenham (UK) and Northampton (US):
Edward Elgar Publishers.
Cassoni, A. and M. Fachola (1997): “Non financial services sector: perspectives in view of
the openness and integration process”, unpublished report for the Ministry of Labor,
Uruguay.
Cox, David (1994): “Some Applied General Equilibrium Estimates of the Impact of a North
American Free Trade Agreement on Canada”, in Francois, J. and C. Shiells:
Modeling Trade Policy. Applied General Equilibrium Assessments of North American
Free Trade, Cambridge University Press.
Croce, H., L. Macedo, and P. Triunfo (2000): “Cost function: an analysis of Uruguayan
commercial banks”, Working Paper 1/00, Department of Economics, School of Social
Sciences, Universidad de la República, Uruguay.
Frenkel, Roberto, and Martín González Rozada (2000): “Liberalización del balance de
pagos. Efectos sobre el crecimiento, el empleo y los ingresos en Argentina –
Segunda parte”. Buenos Aires, CEDES (mimeo).
Ganuza, E., R. Paes de Barros, R. Vos (2002): “Labour Market Adjustment, Poverty and
Inequality during Liberalisation”, in: R. Vos, L. Taylor and R. Paes de Barros (eds.)
Economic Liberalisation, Distribution and Poverty: Latin America in the 1990s,
Cheltenham (UK) and Northampton (US): Edward Elgar Publishers.
Ganuza, E., S. Morley, S. Robinson, and R. Vos (2004): ¿Quién se beneficia del libre
comercio? Promoción de exportaciones y pobreza en América Latina y el Caribe en
los 90. New York, UNDP-Alfaomega Colombiana S.A.
Laens, Silvia and Inés Terra (2000): “Efectos del perfeccionamiento del MERCOSUR sobre
el mercado de trabajo de Uruguay: un ejercicio de simulación usando un modelo
28
CGE”, Revista de Economía, Banco Central del Uruguay, 7 (2), segunda época,
November.
Laens, S., M. Perera (2004). “Uruguay: crecimiento de las exportaciones, pobreza y
distribución del ingreso”, in: E. Ganuza, S. Morley, S. Robinson, R. Vos (eds).
¿Quién se beneficia del libre comercio? Promoción de exportaciones y pobreza en
América Latina y el Caribe en los 90. New York, UNDP-Alfaomega Colombiana S.A.
Lorenzo F, R. Osimani, P. Caputti (1999): “Matriz de Insumo Producto y Contablidad Social
para la economía uruguaya. Año 1995”, Impacto de la apertura comercial del
MERCOSUR sobre el mercado laboral uruguayo, Documento 3. Montevideo,
Programa para el fortalecimiento del Área Social (FAS), Oficina de Planeamiento y
Presupuesto (Inédito).
Oaxaca, R. (1973): “Male-Female Wage Differentials in Urban Labor Markets”, International
Economic Review, 14.
Paes de Barros, R. (1999): “Evaluando el impacto de cambios en la estructura salarial y del
empleo sobre la distribución de renta”, IPEA, Rio de Janeiro (Mimeo).
Paes de Barros, R. and P. Leite (1998): “O Impacto da Liberalizaçao sobre Distribuiçao de
Renda no Brasil”, IPEA, Rio de Janeiro (Mimeo).
Saint Paul, Gilles (1996): Dual labor markets. The MIT Press.
Shapiro, C. and Stiglitz, J. (1984): “Equilibrium unemployment as a worker discipline device”,
American Economic Review, Volume 74, N° 3.
Tansini, R. and Triunfo, P. (1998a): “Technical efficiency and trade openness in the
Uruguayan manufacture sector”, Working Paper 4/98, Department of Economics,
School of Social Sciences, Universidad de la República, Uruguay.
Tansini, R. and P. Triunfo (1998b): “Technical change and productivity in Uruguayan
industrial firms”, Working Paper 12/98, Department of Economics, School of Social
Sciences, Universidad de la República, Uruguay.
Thierfelder, K.E., C.R. Shiells (1997): “Trade and labor market behavior”, in J.F. Francois
and K. A. Reinert (eds), Applied methods for trade policy analysis. A handbook.
Cambridge, Cambridge University Press.
United Nations Development Program (2001): “Human Development in Uruguay Report”.
29
Appendix 1 Wage gap between formal and informal workers
Even if there is no widely accepted and accurate definition of informality, the term is often
used to refer to economic activities that are not illegal but avoid government regulations.
From the labour perspective, workers are considered to be informal when they are not
covered -- in practice -- by labour regulations. These regulations include the different
aspects of labour legislation, taxation and the entitlement to certain benefits such as the paid
sick leave or the retirement pension.
Because of the broad set of aspects covered by labour regulations, it is necessary to choose
an operational definition. For our purpose, workers are considered informal when they have
a job but do not contribute to the social security system. This contribution is the only
regulation that is mandatory to the whole labour force regardless of one’s occupation. In
turn, the contribution entitles workers to receive a pension during retirement. Besides, the
system provides other benefits -- less important in coverage and spending -- to some
contributors during their working life e.g. health benefits, family allowances, pensions for the
widow and children in case of death, among others.
The data
To estimate the wage gap between formal and informal workers, we used the Continuous
Household Survey (CHS) collected by the National Statistics Institute (INE) in 2003. The
CHS is a survey carried out in urban areas that inquires about personal and labour
characteristics (age, sex, marital status, schooling, hours of work, occupation, industry, etc.)
and income received the preceding month classified by sources (wages, pensions, interest
payments, etc).
We restricted the sample to the wage earners and self-employed. This means that we
excluded the following: people who work in a family enterprise without receiving pay, owners
of firms (regardless of size), and members of cooperative units. These groups represent
around 5 percent of the active population.
We also limited the sample to workers of 18 to 59 years old, who represent 10 percent of the
labour force. The bottom border was chosen because there are specific regulations for
30
workers younger than 18 years old (the minimum legal work age is 14). Regarding the top
border, 60 years old is the minimum required retirement age.
In order to classify a worker as formal or informal we used his status of contributor to the
security system for reported information on his main occupation. Accordingly, we worked
with data on the earnings and characteristics of the main job.
The earnings were calculated as the sum of in-cash and in-kind monthly regular labour
income divided by 4.2 (number of weeks in a month), and multiplied by the hours worked in
the preceding week. The monthly regular labour income included: i) the regular earnings
reported in the CHS; ii) the monthly in-the-job health benefits estimated by INE and cited in
the CHS; iii) an estimation of the so-called thirteenth month wage; and iv) an estimation of a
pecuniary benefit received during holidays.
The thirteenth month wage is the right of private and public wage earners to receive an extra
monthly wage equivalent during a year. The CHS reports the receipt of this benefit in the
worker’s main job. Where a positive answer was given, we added an amount equivalent to
1/12 of the reported monthly in-cash regular wage.
Specifically for private sector wage earners, the law establishes a pecuniary benefit to be
received during holidays. However, the CHS does not collect information about this benefit.
To approximate this benefit, we added an amount equivalent to 1/18 of the reported monthly
in-cash regular wage when the worker was a private wage earner and reported to receive a
thirteenth month wage.
We made different estimations of the wage gap between formal and informal workers.
First, we used a very simple econometric model: we regressed by OLS the log hourly wage
on individual and labour characteristics, including a dummy variable that identified if the
worker was formally employed. Let W be the wage of a worker, X its observable
characteristics and F a variable that has value1 when the worker is formal (contributes to the
social security system):
( ) εβ ++= FGXW 1ln1
31
The estimated ‘dummy’ parameter G1 reflects the wage gap between formal and informal
workers.
Next, we used the usual way of decomposing wage differences proposed by Oaxaca (1973)
and Blinder (1973). We divided the sample into two sub-samples, one of formal workers and
another of informal workers, and an earning equation was estimated for each one. Let W be
the wage of a worker, X its observable characteristics and f,i two sub-indices that denote
formality and informality respectively:
( )( ) iiXiiW
ffXffW
εβ
εβ
+=
+=
ln3
ln2
We assume that εj (j=i, f) is an error term with a normal distribution with zero-mean and we
estimate both equations by OLS. Denoting the mean of the variables with a bar and making
some calculations, we can decompose the raw gap between sectors as:
( ) ( ) ( ) ( ) ( ) ( )iffiififififif XXXXXXWW ββββββ ˆˆˆˆˆˆlnln4 '''' −+−=−+−=−
The last of the components reflects the wage difference that is not explained by independent
variables but by the coefficients of the earnings equations. This may be interpreted as the
wage gap valuated in the mean of the formal worker’s characteristics. An analogous
decomposition allows estimating the wage gap as the difference between coefficients but
weighted by the average characteristics of informal workers thus:
( )( )iff
ifi
XG
XG
ββ
ββ
ˆˆ3
ˆˆ2'
'
−=
−=
This estimation ignores the endogeneity of the selection decision of being formal or informal.
We expect unobservable individual characteristics to be correlated with being either formal
or informal (i.e. people with easy access to informal networks or to informal benefits could
have more potential gains in being informal). One strategy to deal with this kind of a problem
consists of estimating a switching regression model.
A latent variable F* defines a variable F that takes value 1 when the worker is formal and 0
when he is informal. The variable F* depends on two different types of characteristics: those
that affect the level of earnings and hence the choice of being formal or informal (X) and
32
those that have a direct effect on this choice (Z). The model is completed with two wage
equations:
The disturbances η are potentially correlated with ωi and ωf. We assume that they have a
trivariate normal distribution and we do a joint estimation using the full-information
maximum-likelihood method. The wage gap between formal and informal workers is
estimated by calculating the predicted difference in earnings. Similar to the OLS estimations,
we estimate the two gaps thus:
( )( )iff
ifi
XG
XG
αα
αα
ˆˆ5
ˆˆ4'
'
−=
−=
Results
The results of the earning equation proposed in equation (1) are reported in column (A) of
Table 1. We controlled personal characteristics (age, education, gender, marital status,
geographical region), the type of occupation (public servants, size of the establishment of
the private wage earners, self-employed who own some property and self-employed who do
not and other labour characteristics (part-time and industry). In columns (B) and (C) we
report the results of the estimation of equations (2) and (3). Finally, the results of the
switching regression model estimations appear in the last columns. The signs of the effect of
the usual explanatory variables included in the earning equation are expected: labour
income increases with education, rises with age at decreasing rates and is higher for married
people and for men.
We report the predicted difference in earnings in Table 2. The five alternative estimations
suggest that the formal workers are highly remunerated.
( )
( )( ) 1 Fln7
0 Fln60F ; 0 *F1F
*F5
=+==+=
=>=++=
ifXWifXW
otherwiseifZX
ffff
iiii
ωαωα
ηπμ
33
Table 1. Results of regression estimates OLS regression estimates Switching regression estimates
Whole sample (A)
Formal workers (B)
Informal workers (C)
Sector allocation*** (D)
Formal workers (E)
Informal workers (F)
0,592 Formal (41.63)**
0,163 0,07 0,182 0,263 0,093 0,162 6 to 8 years of schooling (6.49)** (2.37)* (5.12)** (29.62)** (19.76)** (31.67)**
0,294 0,212 0,296 0,563 0,256 0,247 9 to 11 years of schooling (11.07)** (6.98)** (7.39)** (60.39)** (52.92)** (42.64)**
0,446 0,379 0,421 0,811 0,434 0,343 12 years of schooling (16.47)** (12.31)** (9.84)** (84.04)** (88.70)** (52.96)**
0,662 0,56 0,788 0,977 0,626 0,689 Tertiary level incomplete (21.35)** (16.51)** (12.33)** (78.10)** (116.85)** (75.02)**
0,902 0,805 1,118 1.601 0,894 0,925 Tertiary level complete (30.54)** (24.59)** (13.28)** (127.72)** (170.94)** (75.36)**
0,051 0,059 0,048 0,055 0,063 0,043 Age (14.93)** (16.46)** (7.68)** (39.09)** (110.35)** (46.43)**
-0,05 -0,057 -0,053 -0,044 -0,061 -0,048 Age squared (/100) (11.54)** (12.50)** (6.36)** (24.75)** (85.66)** (40.83)**
0,119 0,093 0,149 0,22 0,108 0,125 Civil status (Married=1) (11.28)** (8.33)** (7.33)** (38.10)** (61.73)** (39.91)**
-0,217 -0,209 -0,243 0,038 -0,209 -0,247 Gender (Female=1) (20.14)** (19.06)** (9.94)** (6.93)** (125.31)** (70.99)**
-0,06 -0,127 0,016 0,234 -0,115 -0,005 Agriculture (2.11)* (3.67)** -0,35 (23.80)** (23.99)** -0,68
0,352 0,262 -0,19 0,542 0,276 -0,296 Electricity, water & gas (9.79)** (7.35)** -0,42 (8.80)** (41.80)** (3.17)**
0,128 0,037 0,194 -0,312 0,011 0,206 Construction (4.85)** -1,36 (4.81)** (30.42)** (2.16)* (34.97)**
-0,055 -0,068 -0,011 0,254 -0,05 -0,034 Commerce (3.32)** (3.96)** -0,35 (39.83)** (17.29)** (7.26)**
0,076 0,055 0,025 0,395 0,079 -0,026 Transport (3.38)** (2.46)* -0,44 (40.92)** (22.12)** (3.23)**
0,241 0,234 0,163 0,556 0,273 0,101 Finance (10.37)** (9.27)** (3.37)** (62.69)** (80.96)** (14.17)**
0,102 0,007 0,248 0,084 0,018 0,24 Services (6.08)** -0,39 (7.53)** (12.19)** (6.27)** (48.38)**
0,416 0,395 0,427 -0,863 0,334 0,504 Part time (less than 30 hours=1) (28.03)** (20.86)** (20.12)** (163.92)** (120.11)** (129.86)**
0,209 0,169 0,271 0,196 0,181 0,251 Region (Montevideo=1) (21.09)** (16.30)** (13.83)** (47.65)** (111.85)** (84.36)**
0,231 -0,007 0,167 3.382 0,32 -0,29 Public servant (8.09)** -0,1 -1,49 (258.43)** (33.93)** (13.65)**
0,137 -0,08 0,129 0,7 0,052 0,094 Self-employed with property (4.88)** -1,09 (3.92)** (72.13)** (6.40)** (20.92)**
0,117 -0,21 0,132 1.352 -0,034 0,032 Private micro-enterprise (< 5) (4.56)** (2.95)** (4.46)** (141.40)** (4.03)** (5.95)**
0,163 -0,163 0,25 1.865 0,084 0,08 Private little enterprise (5-9) (5.76)** (2.29)* (6.45)** (178.31)** (9.43)** (9.74)**
0,289 0,027 0,231 2.619 0,33 -0,067 Private other size (>9) (10.73)** -0,38 (5.84)** (271.86)** (36.27)** (6.01)**
0,022 Household head (3.37)**
0,031 Household head’s spouse (3.82)**
0,058 School attendance (attendance=1) (6.33)**
-0,306 Retirement pension (recipient =1) (31.14)**
-104 x 4.7 Household income (log) (-0.53)
1,057 1,87 1,068 -3,613 1,362 1,219 Constant (15.24)** (18.91)** (9.09)** (123.27)** (84.85)** (67.54)** Observations 17767 11450 6317 R-squared 0,48 0,4 0,23 * Denotes significance at 5%; ** denotes significance at 1%; *** Correlations: η and ωf = 0,41314*; η and ωI = -0,30800*
34
Table 2. Estimated mean difference in earnings between formal and informal workers (log Wf – log Wi) Raw gap 0,85 Estimated gap G1 0,59 G2 0,65 G3 0,60 G4 0,52
35
Appendix 2: The CGE model
Equations
The equations of the CGE model are presented in this appendix. Three versions of
the model were specified: perfect competition in the labour market, efficiency wages for non
skilled workers, and the wage curve. Lower fonts indicate endogenous variables, capital
fonts refer to exogenous variables, and Greek letters indicate parameters. The subscripts i, j
refer to sectors, the subscripts z, t refer to geographic zones and the subscripts f refer to
representative households grouped according to income levels as follows:
i, j = {1, 2, …, J}
z = Uruguay (u), Argentina (a), Brazil (b), rest of the world (r)
t = a, b, r
f=(f1,f2,f3,f4,f5,f6,f7,f8,f9,f10)
k=( f1,f2,f3,f4,f5,f6,f7,f8,f9,f10,g)
Demand Structure
Demand functions are derived from a Cobb Douglas utility function which is an
increasing function of consumption of composite goods that combines different varieties of
differentiated goods. In turn, the sub-utility functions follow an Armington specification (1969)
in perfect competition sectors. In the perfectly competitive sectors, goods are differentiated
by geographic origin.
Consumers maximize a Cobb Douglas utility function subject to their budget
constraint. As such, demand for each good is stated thus:
i
fffifif pf
msavtdyc
)1)(1(.
−−= μ (1)
where cif is the demand for a composite final good i (differentiated by geographic origin), Yf
is total income of a representative household f in Uruguay, tdf is direct tax rate, msavf is
marginal propensity to save, and pfi is the composite price index. This index can be written
as:
( ))1/(1
1i
ii
zzizii ppf
φφφλ
−− ⎟
⎠
⎞⎜⎝
⎛= ∑ , (2)
being λzi the share parameter in the Armington function, Фi the elasticity of substitution
between goods from different origin, and pzi the market price of good i from market z.
36
Investment demand of good i is a fixed share of total investment:
(3)
I being total investment.
Final demand of a differentiated good i produced in country z by a representative
household f is:
kii
zizizik c
pfpd
i
i ..φ
φλ−
⎟⎟⎠
⎞⎜⎜⎝
⎛= (4)
where dzih is the final domestic demand of the ia<institution f.
The export demand for a representative domestic firm is a decreasing function of the
export price:
0 . ..
i
i
iz iz tiz
zi
e p ReER pd
η
η
−
−= (5)
where eiz is the demand for a variety of the differentiated good i in market z, piz is the export
price from Uruguay, pdzi is the domestic price index of good i in market z, R is the real
income of the partner z, ER is the exchange rate, and eoiz is a parameter.
Production Each sector combines primary factors and intermediate inputs following a Cobb-
Douglas production function. The value added is a nested CES production function
combining skilled labour, unskilled labour, and capital.
Cost
Total variable cost is derived from a Cobb-Douglas constant returns to scale
production function. The variable unit cost is:
( )( ) ∏∑+= −
jjiiiii
jij
ji vitindvcv ααω .1 1 (6)
where vi is the variable unit cost, vci is the value added cost and viij is the composite price of
intermediate inputs. αij is the distribution parameter of a Cobb-Douglas production function,
and ϖi is a parameter.
Value added is a combination of labor and capital specified as a CES. Thus, vci is:
( )[ ] )1/(1)1()1( ..1 iiiiiiiii wrvc σσσσσ δδ −−− +−= (7)
where ri y wi,, are the rental rate of capital and the average wage. δ is distribution
iiinviinv pf
Ic μ=
37
parameters of the CES function for value added, while σi is the elasticity of substitution
between capital and labour.
As the model considers two types of labour, the average wage is a combination of
skilled and unskilled wage. It is assumed that skilled and unskilled labours are combined
following a CES function, so the average wage is:
( ) ( )[ ] )1/(111 ...1.1 iiiiiiii
ili wswdwuw
θθθθθ ξξϕ
−−− +−= (8)
where wli is the average wage, wu y wsi are the unskilled and the skilled wage, respectively,
ξ y ϕ are the distribution and scale parameters, and θi is the elasticity of substitution between
skilled and unskilled labour.
The efficiency wage is endogenous. It is assumed that the workers caught from
(What do you mean by this?) the efficiency wages sectors go to the informal sector, where
the labour market is competitive and wage premiums are absent. To model the efficiency
wage premium we follow Thierfelder and Shiells (1997):
( )( )
( ) ⎟⎠
⎞⎜⎝
⎛−−
++
−=
−
∑ormali
i
i
luLUDD
ULSDDD
rdwd
wd
inf12(
1)12(
.1
m
κκ (9)
where κ is the utility of shirking, rd is the discount rate, D! is the probability that no-
shirking workers will be falsely accused and fired from the efficiency wage sector, D2 is the
probability to be caught shirking and therefore fired, S is the rate of quitting the efficiency
wage sector. Other specifications of the model do not consider situations when a worker is
fired from the efficiency wage sector and remains unemployed. The estimation of the wage
curve will be used to calibrate the parameters.
The intermediate inputs are differentiated by geographic origin with an Armington
formulation. The composite price of intermediates is:
( ))1/(1
1..j
jj
zzjzjiji pvi
φφφγ
−− ⎟
⎠
⎞⎜⎝
⎛= ∑ (10)
where pzj is the price in the local market of input j used in sector i from each zone, γzji is the
CES distribution parameter, and φj is the elasticity of substitution between goods from
different origins.
Input and factor demand by firm
38
Firms maximize their profits so demand for intermediate inputs and value added
(labour and capital) in each sector is obtained from their maximization program:
j
jizji
zj
ji
iijizji vi
pvi
qvx
φ
γα
−
⎟⎟⎠
⎞⎜⎜⎝
⎛=
..
(11)
where xzji is the demand for input j coming from country z and used by sector i for each firm
in sector i. It is a decreasing function of the input price.
Valued added demand is a decreasing function of the value added cost and
increasing function of the unitary cost and output in each sector:
( )ii
iiii tindvc
vqvva
+=
1α (12)
Factor demand is a decreasing function of their return rate and is an increasing
function of value added and its price:
iifi
ii va
vcw
fdi
..
σ
δ
−
⎟⎟⎠
⎞⎜⎜⎝
⎛= (13)
Finally, the skilled and unskilled labour demand equations are the following:
liii
fii fd
wtfacws
lsi
..
1(.
θ
ξ
−
⎟⎟⎠
⎞⎜⎜⎝
⎛ += (14)
liii
fii fd
wtfacwdwu
lui
.)1(
)1(. θ
ξ
−
⎟⎟⎠
⎞⎜⎜⎝
⎛−
+= (15)
Domestic pricing
In the perfect competitive sectors, the equilibrium price of output is equal to its
variable unit cost (vi ):
( )iiui texvp += 1 when i= competitive sectors (16)
where the lower case “u” refers to Uruguay. The firms charge the same price in domestic
and foreign markets.
General Equilibrium
Public services fix prices, wages, and employment whereas production level and
capital demand is endogenous.
Income of the households is endogenous and is the sum of the returns to factors of
production and transfers from the government:
39
lg)...( wgtrfrkwly fiiii
iif ++++=∑ (17)
Government income is the sum of the receipts of tariff collection, indirect taxes and
profits from public firms:
∑ ∑ ∑∑∑∑ ⎟⎟⎠
⎞⎜⎜⎝
⎛+++++=
i z zzj
jzjzjizjuizizizizii
iiiii
iiig pnxnpndtindfrkwly ...).()...( ττπ (18)
Government expenditure is the sum of household transfers, public wages and
government consumption:
lgwgpdtrGE zizigf
f ∑∑ ++= (19)
where GE is the government expenditure, d is the government consumption of good I, which
is a fixed coefficient, wg is the public wage and lg is public employment, both fixed.
Government savings is the difference between government income and expenditure:
GEySG G −= (20)
and assumed to be constant.
The equilibrium conditions in the labour market are:
iiii nfslsLS .+= (21) where LSi is the supply of skilled labour and
∑ +=i
iii nfuluLU ).( (22)
where LU is the supply of unskilled labour. Both variables are exogenous.
The equilibrium equation for capital is:
iiii nfkkK .+= (23) where Ki is capital supply (exogenous).
When factors are assumed to be sector specific there is one equilibrium condition for
each factor and sector, but when factors are assumed perfectly mobile there is only one
equation for each factor.
The equilibrium conditions in the goods market require that supply equals demand in
each sector:
∑ ∑++=j t
ituijuii exdq (24)
Finally, the external equilibrium is:
BpxnpdERpe tji t i j t
tjiuiZItii t
uiit =−−∑∑ ∑ ∑∑∑∑ ... (25)
In all the simulations B is fixed in terms of the numerary.